GREAT GRAPHS

YOUR GUIDE TO GREAT GRAPHS

ANN K. EMERY

WWW.ANNKEMERY.COM

ANALYZE YOUR AUDIENCE During the planning phase, discuss these questions with your teammates: ✓ Who’s my audience? ✓ Are my audiences technical or non-technical? ✓ What’s worth visualizing? ✓ How many points in time do our viewers need to see? ✓ How many decimal places are useful? ✓ What types of comparison data are available? ✓ Are viewers expecting a story? ✓ How will your completed chart ultimately be shared? Within the body of a longer report? As the star of a one-page handout? ✓ Will your graphs be shared in-person, or read on-screen? ✓ How often will this information be shared? Once? Monthly? Quarterly? Annually? ✓ Will the graphs be printed in full color or photocopied in black and white? ✓ Will the page layout be tall and skinny (reports) or short and fat (slideshows)? ✓ What are the paper dimensions and how many pages are needed (e.g., 8.5 x 11, 11 x 17)? ✓ Can the graphs live inside Excel, or will they created within PowerPoint, Word, or another program? If your viewers are expecting a story, then use saturation, titles, and/or annotations to interpret the numbers. Number of studies funded each year

We’re funding more studies than before Beginning in 2013, we set aside new funding to measure the effectiveness of our work.

40

40

30

30

20

20

10

10

0

0 '08

'15

GREAT GRAPHS: THE WORKSHOP WITH ANN K. EMERY

2013: New policy introduced

'08

'15

1

CHOOSE THE RIGHT CHART Sketching

Heat Tables

Data Bars

Spark Lines

Spark Bars

Thank me later

Color-coding

Within-cell bars

Mini trendlines

Mini columns

A B C

20 40 60

'16 A 20 B 40 C 60

A 20 B 40 C 60

'17 55 35 20

'16 A 20 B 40 C 60

'17 55 35 20

Pie/Donut

Stacked

Diverging

Waffle

Icon Array

2-3 slices max!

Pie alternative

Focus on poles

aka Square pies

aka Pictographs

Bar/Column

Lollipop

Clustered

Dot

Multiples

The trusty steed

Minimalist bar

Two pairs, max!

Minimalist cluster

Disaggregated

Area

Nested Area

Overlapping

Target Line

Comparisons based on area

aka Tree

When every bar has a different goal

When every bar has the same goal


PROGRESS

COMPARISONS

PART : WHOLE

EXPLORATORY

Will you use a bar chart, a pie chart, or something else altogether? Sketch a couple options and discuss the pros and cons with a colleague. See more at www.annkemery.com/essentials.

2

Tile or Hex

Tile Trendline

Network

aka Heat Map

Equal-sized squares/hexagons

Geographical patterns over time

Connections between people

3+ POINTS IN TIME

BEFORE/AFTER

DISTRIBUTION

MAPS

Choropleth

Histogram

Pyramid

Scatter

Bubble

Box Plot

Ordinal variables (e.g., age ranges)

Ordinal & nominal (ages & gender)

Correlations (x & y)

Correlations (x, y, & z)

Min, Q1 median, Q3, & max

Slope

Dot

Stacked

Waterfall

Deviation

Focus on the slope of the line

Focus on distance between the dots

Part-to-whole

How parts add up to the whole

Just display the # or % differences

Line

Multiples

Stacked Bar

Stacked Area

The classic choice

Separate the spaghetti

Part-to-whole

Part-to-whole

Target Line

Estimation

Fan

Calendar

Timeline

Goal is light

Guess is dotted

Guess is shaded

Milestones

Try words & icons


3

SELECT A SOFTWARE PROGRAM Excel, Tableau, and R are the some of the most popular tools on the market. These multitasking software programs allow you to both crunch your numbers and design your graphs, and their default graphing settings— while imperfect—are highly customizable with just a few clicks.

Tool

Pros

Cons

Get Started

Excel

Virtually free.

Interactive graphs possible but require advanced skills.

Watch tutorials on timesaving spreadsheet strategies and graphing basics: annkemery.com/excel

Minimal learning curve for standard graphs (bar, line, scatter, etc.). Additional graph types possible through workarounds (dot, slope, small multiples, etc.).

Not ideal for real-time analysis of big datasets. Typically slows down with 20,000+ rows of data.

Coworkers can easily view your graphs since most people already have Excel.

Tableau

Built-in library of charts contains more layouts than Excel (e.g., maps). Drag-and-drop design allows users to explore which chart type might work best in just a few clicks. Static or interactive visualizations.

R

Free. Vibrant learning community; users develop tutorials and freely share code with others.

Free version: Uploaded data is publicly available; do not upload sensitive information such as names or addresses. Paid version: Can be cost-prohibitive.

Download a trial version: tableau.com Watch screencasts from Tableau staff: youtube.com/tableausoft ware

Coworkers need to download Tableau Reader to view your graphs. Prior coding experience recommended.

Download R: r-project.org

Expect a learning curve.

Follow the step-by-step learning plan outlined at annkemery.com/portfolio /r

Ideal for large datasets. Static or interactive visualizations.


4

DECLUTTER Once you’ve built your graph in the software program of your choosing, it’s time to roll up your sleeves and make intentional edits. Good data visualization is all about creating contrast. First, declutter by deleting as much ink as possible. Then, then sharpen your desired message with titles, subtitles, annotations, labels, and color.

Remove borders…



…from the graph as a whole. …from segments within the graph (e.g., stacked bar charts get outlined in white). …from tables.



Remove tick marks.



Remove 3D.



Remove bevels.



Remove leader lines.



Remove background fills/decorations.



Remove decorative clip art. Optional: Use icons to aid interpretation.

Label directly.



Delete the legend and place labels directly beside or on top of the data.

Label sparingly.



Choose axis labels or numeric labels, but not both. In this example, I intentionally removed the border, the vertical grid lines, and the horizontal grid lines. The xand y-axes are still there, but I lightened the remaining lines (from black to gray ink). 100

100

80

80

60

60

40

40

20

20

0

0 0

20

40

60

80

100


0

20

40

60

80

100

5

CLARIFY WITH COLOR Color can enhance branding and guide your viewers to the most important pieces of the graph.



Brand visuals with custom colors. Consult a style guide or find RGB codes with eyedroppers (e.g., instant-eyedropper.com). In Excel, select Shape Fill (the paint can)  More Fill Colors  Custom.



Ensure legibility in grayscale. Preview a draft in grayscale. Is there enough contrast between the dark gray and light gray?



Ensure legibility for people with color vision deficiencies. Avoid red/green combinations. Instead, use orange/green or red/blue. Test drafts at www.color-blindness.com.



Color-code by category or chapter (e.g., each chapter in a different color from your logo).



Color-code by variable. • • • •



Binary variables: Select one hue from your logo. The darker hue represents presence and the lighter hue represents absence. Sequential: Select one hue from your logo in which darker hues represent presence. Diverging: Select two hues from your logo. The darkest version of each hue will represent the outer poles. Categorical/nominal: Select a different hue for each category.

Saturate to draw attention. The important sections of the graph will be one of the darker hues from your logo and the least important sections will be gray (or, a less saturated version of that same hue from your logo).

This seal’s blue (RGB 3, 88, 150) is similar, but not identical, to your computer’s default color scheme. We’ll still need to manually adjust the RGB codes.

A

50

B

40

C


18 45

D E

15

30 15

9

7

6

6

CLARIFY WITH TEXT Your graph’s title, subtitle, and/or annotations can work together to strengthen your message.



Match the graph’s font to branding guidelines. Consult a style guide or identify fonts with www.whatthefont.com.



State the story in the title. 6 to 12 words.



Add a subtitle to written products. 1 or 2 sentences.



Left-align titles and subtitles.



Annotate.



Measure text readability with www.readable.io.



Make sure text is horizontal rather than vertical or diagonal.



Establish a text size hierarchy. The most important information should be large and the least important information should be small.

Before, the states were twisted diagonally, which means your poor viewers would have to tilt their heads to read your graph. After we rotated the graph, our viewers can read the horizontal text much faster. 95 Virginia

80 70 60

95

Maryland

55

80

New York 40

32

30

70

New Jersey 25

60

South Dakota

55

Montana Oregon

32

Minnesota

30

Ohio


40

25

7